www.nature.com/scientificreports
OPEN
Received: 15 March 2017 Accepted: 9 June 2017 Published: xx xx xxxx
Distinct contributions by frontal and parietal cortices support working memory Wayne E. Mackey 1 & Clayton E. Curtis1,2 Although subregions of frontal and parietal cortex both contribute and coordinate to support working memory (WM) functions, their distinct contributions remain elusive. Here, we demonstrate that perturbations to topographically organized human frontal and parietal cortex during WM maintenance cause distinct but systematic distortions in WM. The nature of these distortions supports theories positing that parietal cortex mainly codes for retrospective sensory information, while frontal cortex codes for prospective action. Working memory (WM) bridges perception and action over brief periods of time1 and acts as a critical building block for high-level cognitions2–4. While a large network of brain areas support WM5, past research demonstrates that within portions of frontal and parietal cortex, specifically, the activities of neurons persist during WM retention intervals6–8. Moreover, chronic lesions to both of these areas impair WM ability9–12. The patterns of neural activity during WM and the consequences of damage are so strikingly similar, it remains unknown what distinct contributions, if any, the frontal and parietal cortices provide to support WM. We addressed this key limitation by testing the hypothesis that the nature of the maintained information differs in each area. Based on theories of sensorimotor dynamics1, 13, we predict that the parietal cortex largely maintains representations of past sensory information, while the frontal cortex largely maintains representations of future plans. We test this hypothesis by combining computational neuroimaging and transcranial magnetic stimulation (TMS) to transiently disrupt activity in topographically defined subregions of frontal and parietal cortices while subjects actively maintained information in WM. We used functional magnetic resonance imaging (fMRI) and nonlinear population receptive field (pRF) mapping (Fig. 1a,b)14, 15 to identify potential stimulation sites in individual subjects(Fig. S1-S3), including four visual field maps in parietal cortex (IPS0, IPS1, IPS2, and IPS3) and two visual field maps in frontal cortex (SPCS, IPCS) (Fig. 1c,d). We targeted the superior precentral sulcus (sPCS) in frontal cortex and the third intraparietal sulcus area (IPS2) in parietal cortex because those topographic maps are the putative human homologues of the monkey frontal eye field (FEF) and lateral intraparietal area (LIP), respectively. In both species, these areas exhibit delay-period activity during WM tasks6, 8, 16, and lesions to these areas in both species cause impairments in WM10–12, 17, 18. Additionally, we targeted the dorsolateral prefrontal cortex (PFC), an area of frontal cortex also shown to be critical for WM in nonhuman primates9, but whose role in human WM is controversial12. Consistent with previous studies, we found no coherent topographic map of visual space in dorsolateral PFC15, 19, 20, and therefore used anatomical landmarks for criteria to identify dorsolateral PFC in individual subjects(Fig. S3)21. We measured the degree to which disrupting neural activity with TMS in each target area affected WM performance, which was assayed by the accuracy of memory-guided saccades (Fig. 1e). Subjects generated saccades following a memory retention interval to the location of a target briefly flashed before the delay. During TMS sessions, but not control sessions, we applied short trains of patterned TMS in the middle of the delay period on each trial (Fig. 1f). Subjects typically make an initial ballistic eye movement called a memory-guided saccade (MGS) towards the remembered target, followed by small and sometimes multiple corrective eye movements that usually bring the gaze closer in alignment with the remembered target (Fig. 1g). Therefore, we measured the accuracies of both the initial MGS and the final eye position (FEP) on each trial, as these two measures are likely to be sensitive to distinct components of WM. For example, the developmental trajectory of MGS accuracy leads that of FEP accuracy by about five years, and does so in parallel with the earlier maturation of motor systems compared to the later maturation of cognitive systems22. MGS and FEP show different patterns of sensorimotor 1 Department of Psychology, New York University, New York University, 6 Washington Place, New York, NY, 10003, USA. 2Center for Neural Science, New York University, New York University, 6 Washington Place, New York, NY, 10003, USA. Correspondence and requests for materials should be addressed to C.E.C. (email: [email protected])
Scientific Reports | 7: 6188 | DOI:10.1038/s41598-017-06293-x
1
www.nature.com/scientificreports/
Figure 1. Experimental procedures. (a) Discrimination task used for topographic mapping. Subjects fixated centrally while covertly attending to a bar consisting of three apertures of moving dots that swept across the screen. Subjects pressed a button to indicate which flanking aperture (left or right; above or below) contained dots moving in the same direction as the center sample aperture. Staircases adjusted dot motion coherence in the flanking apertures to constantly tax attention. White outlines around each aperture were not visible to the subject, but are shown here for clarity. (b) Nonlinear population receptive field (pRF) modeling schematic. Stimulus sequences from the discrimination task were converted into 2D binary contrast apertures and projected onto an isotropic 2D Gaussian that represented a predicted pRF. Static nonlinearity was applied to account for compressive spatial summation. (c) Stimulation sites projected onto the inflated cortical surface of a sample subject. Color indicates the best-fit phase angle by the pRF model. sPCS and IPS2 stimulation sites were localized by our mapping procedure, and dorsolateral PFC stimulation targets were defined by individual subject anatomical landmarks. (d) Visual field map coverage in IPS2 and sPCS. While visual field maps in both regions primarily represent the contralateral hemifield, they also represent portions of the ipsilateral hemifield. (e) Memory-guided saccade task. While fixating, subjects maintained the position of a brief visual target over a memory delay, and then made a saccade to the remembered target. The correct target location was again presented for feedback. Dotted circles depict gaze, but were not visible to subjects. (f) Temporal structure of the experimental task and application of TMS. In order to disrupt the maintenance of information, we applied a short burst of patterned TMS to IPS2, sPCS, or dorsolateral PFC during the middle of the delay period on each trial. (g) An example horizontal eye position trace shows two distinct types of memory error relative to the true target location: the endpoint of an initial memory-guided saccade (MGS), and the final eye position (FEP) following quick corrective saccades (in this example only one) prior to the re-presentation of the target feedback (gray line).
adaptation to perturbations in the veracity of visual feedback during visually-guided and memory-guided saccades23. Moreover, patients with schizophrenia24 and patients with resections of the frontal precentral sulcus12 have selective deficits in MGS, but not FEP accuracy. Therefore, MGS and FEP accuracy may index different components of WM. Specifically, the accuracy of the initial MGS may index the quality of the prospective movement, in this case the saccade plan. The FEP, on the other hand, may be an indicator of the fidelity of retrospective sensory information. These two WM codes differ in the nature of what is maintained in memory, a future action plan or a past sensory event.
Results
TMS disruption in frontal and parietal cortex caused distinct WM impairments (Fig. 2; Fig. S4-S6). Although disruption to sPCS caused increased errors in MGS to the contralateral visual field (p = 0.005; TMS worsened MGS in every subject), and to a lesser extent to the ipsilateral visual field (p = 0.02), it had no effect on the accuracy of FEP (contralateral p = 0.54, ipsilateral p = 0.74). Disruption to IPS2 caused increased errors in FEP in the contralateral visual field (p = 0.0005; TMS worsened FEP in every subject), but not ipsilateral visual field (p = 0.24). Similar to sPCS, IPS2 disruption also increased MGS errors to the contralateral visual field (contralateral p = 0.04; ipsilateral Scientific Reports | 7: 6188 | DOI:10.1038/s41598-017-06293-x
2
www.nature.com/scientificreports/
Figure 2. Group average and individual working memory performance. (a) Compared to a baseline no TMS condition, sPCS TMS caused an increase in mean MGS errors especially in the visual field contralateral to the hemisphere of TMS, but not FEP errors. IPS2 TMS impaired mean FEP errors in the contralateral visual field. Dorsolateral PFC TMS caused no observable impairments. †p
OPEN
Received: 15 March 2017 Accepted: 9 June 2017 Published: xx xx xxxx
Distinct contributions by frontal and parietal cortices support working memory Wayne E. Mackey 1 & Clayton E. Curtis1,2 Although subregions of frontal and parietal cortex both contribute and coordinate to support working memory (WM) functions, their distinct contributions remain elusive. Here, we demonstrate that perturbations to topographically organized human frontal and parietal cortex during WM maintenance cause distinct but systematic distortions in WM. The nature of these distortions supports theories positing that parietal cortex mainly codes for retrospective sensory information, while frontal cortex codes for prospective action. Working memory (WM) bridges perception and action over brief periods of time1 and acts as a critical building block for high-level cognitions2–4. While a large network of brain areas support WM5, past research demonstrates that within portions of frontal and parietal cortex, specifically, the activities of neurons persist during WM retention intervals6–8. Moreover, chronic lesions to both of these areas impair WM ability9–12. The patterns of neural activity during WM and the consequences of damage are so strikingly similar, it remains unknown what distinct contributions, if any, the frontal and parietal cortices provide to support WM. We addressed this key limitation by testing the hypothesis that the nature of the maintained information differs in each area. Based on theories of sensorimotor dynamics1, 13, we predict that the parietal cortex largely maintains representations of past sensory information, while the frontal cortex largely maintains representations of future plans. We test this hypothesis by combining computational neuroimaging and transcranial magnetic stimulation (TMS) to transiently disrupt activity in topographically defined subregions of frontal and parietal cortices while subjects actively maintained information in WM. We used functional magnetic resonance imaging (fMRI) and nonlinear population receptive field (pRF) mapping (Fig. 1a,b)14, 15 to identify potential stimulation sites in individual subjects(Fig. S1-S3), including four visual field maps in parietal cortex (IPS0, IPS1, IPS2, and IPS3) and two visual field maps in frontal cortex (SPCS, IPCS) (Fig. 1c,d). We targeted the superior precentral sulcus (sPCS) in frontal cortex and the third intraparietal sulcus area (IPS2) in parietal cortex because those topographic maps are the putative human homologues of the monkey frontal eye field (FEF) and lateral intraparietal area (LIP), respectively. In both species, these areas exhibit delay-period activity during WM tasks6, 8, 16, and lesions to these areas in both species cause impairments in WM10–12, 17, 18. Additionally, we targeted the dorsolateral prefrontal cortex (PFC), an area of frontal cortex also shown to be critical for WM in nonhuman primates9, but whose role in human WM is controversial12. Consistent with previous studies, we found no coherent topographic map of visual space in dorsolateral PFC15, 19, 20, and therefore used anatomical landmarks for criteria to identify dorsolateral PFC in individual subjects(Fig. S3)21. We measured the degree to which disrupting neural activity with TMS in each target area affected WM performance, which was assayed by the accuracy of memory-guided saccades (Fig. 1e). Subjects generated saccades following a memory retention interval to the location of a target briefly flashed before the delay. During TMS sessions, but not control sessions, we applied short trains of patterned TMS in the middle of the delay period on each trial (Fig. 1f). Subjects typically make an initial ballistic eye movement called a memory-guided saccade (MGS) towards the remembered target, followed by small and sometimes multiple corrective eye movements that usually bring the gaze closer in alignment with the remembered target (Fig. 1g). Therefore, we measured the accuracies of both the initial MGS and the final eye position (FEP) on each trial, as these two measures are likely to be sensitive to distinct components of WM. For example, the developmental trajectory of MGS accuracy leads that of FEP accuracy by about five years, and does so in parallel with the earlier maturation of motor systems compared to the later maturation of cognitive systems22. MGS and FEP show different patterns of sensorimotor 1 Department of Psychology, New York University, New York University, 6 Washington Place, New York, NY, 10003, USA. 2Center for Neural Science, New York University, New York University, 6 Washington Place, New York, NY, 10003, USA. Correspondence and requests for materials should be addressed to C.E.C. (email: [email protected])
Scientific Reports | 7: 6188 | DOI:10.1038/s41598-017-06293-x
1
www.nature.com/scientificreports/
Figure 1. Experimental procedures. (a) Discrimination task used for topographic mapping. Subjects fixated centrally while covertly attending to a bar consisting of three apertures of moving dots that swept across the screen. Subjects pressed a button to indicate which flanking aperture (left or right; above or below) contained dots moving in the same direction as the center sample aperture. Staircases adjusted dot motion coherence in the flanking apertures to constantly tax attention. White outlines around each aperture were not visible to the subject, but are shown here for clarity. (b) Nonlinear population receptive field (pRF) modeling schematic. Stimulus sequences from the discrimination task were converted into 2D binary contrast apertures and projected onto an isotropic 2D Gaussian that represented a predicted pRF. Static nonlinearity was applied to account for compressive spatial summation. (c) Stimulation sites projected onto the inflated cortical surface of a sample subject. Color indicates the best-fit phase angle by the pRF model. sPCS and IPS2 stimulation sites were localized by our mapping procedure, and dorsolateral PFC stimulation targets were defined by individual subject anatomical landmarks. (d) Visual field map coverage in IPS2 and sPCS. While visual field maps in both regions primarily represent the contralateral hemifield, they also represent portions of the ipsilateral hemifield. (e) Memory-guided saccade task. While fixating, subjects maintained the position of a brief visual target over a memory delay, and then made a saccade to the remembered target. The correct target location was again presented for feedback. Dotted circles depict gaze, but were not visible to subjects. (f) Temporal structure of the experimental task and application of TMS. In order to disrupt the maintenance of information, we applied a short burst of patterned TMS to IPS2, sPCS, or dorsolateral PFC during the middle of the delay period on each trial. (g) An example horizontal eye position trace shows two distinct types of memory error relative to the true target location: the endpoint of an initial memory-guided saccade (MGS), and the final eye position (FEP) following quick corrective saccades (in this example only one) prior to the re-presentation of the target feedback (gray line).
adaptation to perturbations in the veracity of visual feedback during visually-guided and memory-guided saccades23. Moreover, patients with schizophrenia24 and patients with resections of the frontal precentral sulcus12 have selective deficits in MGS, but not FEP accuracy. Therefore, MGS and FEP accuracy may index different components of WM. Specifically, the accuracy of the initial MGS may index the quality of the prospective movement, in this case the saccade plan. The FEP, on the other hand, may be an indicator of the fidelity of retrospective sensory information. These two WM codes differ in the nature of what is maintained in memory, a future action plan or a past sensory event.
Results
TMS disruption in frontal and parietal cortex caused distinct WM impairments (Fig. 2; Fig. S4-S6). Although disruption to sPCS caused increased errors in MGS to the contralateral visual field (p = 0.005; TMS worsened MGS in every subject), and to a lesser extent to the ipsilateral visual field (p = 0.02), it had no effect on the accuracy of FEP (contralateral p = 0.54, ipsilateral p = 0.74). Disruption to IPS2 caused increased errors in FEP in the contralateral visual field (p = 0.0005; TMS worsened FEP in every subject), but not ipsilateral visual field (p = 0.24). Similar to sPCS, IPS2 disruption also increased MGS errors to the contralateral visual field (contralateral p = 0.04; ipsilateral Scientific Reports | 7: 6188 | DOI:10.1038/s41598-017-06293-x
2
www.nature.com/scientificreports/
Figure 2. Group average and individual working memory performance. (a) Compared to a baseline no TMS condition, sPCS TMS caused an increase in mean MGS errors especially in the visual field contralateral to the hemisphere of TMS, but not FEP errors. IPS2 TMS impaired mean FEP errors in the contralateral visual field. Dorsolateral PFC TMS caused no observable impairments. †p
